/// case study 033
AI Platform
Project 033

StudioOS

StudioOS is two halves over one database. The first is an analytics command centre that pulls App Store Connect, RevenueCat, ad-platform and social numbers into a single daily read, attributing published creative back to the campaign that produced it by content matching rather than manual tagging. The second is a generative pipeline: it learns which formats perform from a corpus of real data, writes original multi-slide creative on a schedule, generates every image through diffusion models behind an automated judge that holds each result to its written brief, retries or escalates what fails, and meters every cent against a hard daily ceiling. A human approval gate sits in front of anything that ships.

Private repo · Source on request
1,550
Unit tests
~$0.30
Per carousel
Cockpit + operator portal
Surface
AI Platform
Category
The Problem

What Wasn't Working

Marketing a consumer app organically means publishing original creative every single day. Recycling one source hits a hard ceiling fast, and producing it by hand does not scale to a person working alone.

The Solution

How I Fixed It

Learn the format from data, generate distinct copy and distinct photography for every slot, check each image and each line automatically against the brief and the content rules, meter the spend against a ceiling, and keep a human approval gate before anything publishes.

Stack

Technologies Used

Next.js 16
Supabase
TypeScript
Vercel
Gemini
Image models
Apify
Results

Key Outcomes

✓
Original creative generated on a schedule, no duplicated source material
✓
Automated picture judging against a written brief, with retries and cost ceilings
✓
Published-content attribution by content matching, no manual tagging
✓
1,550 unit tests, typechecked and linted on every change

Want something like this?

Let's build it. I ship fast and I ship clean.